Without doubt, big data have caught the attention of many businesses across the globe. The availability and richness of data available in public data sources have immensely grown to such levels, that it cannot be ignored. Efforts have been made to analyze and make sense of structured sources. However, unstructured data which comes in many shapes and forms is yet to be utilized to full scale. One type unstructured data sources in text available on social media sites. Just like the viral nature of word of mouth, the digital shapes present eWOM.
An eWOM is a form of communication defined as “a statement made by a potential, actual or former customer about a product or company which is made available to people via the internet”.
eWOMS are less personal, they are emotional and reflect the subconscious opinion. Only 20% of the tweets mentioning a company’s brand contains sentiment, according to Jansen et al. (2009) study. As a result, a great deal of data is required to get a valid sentiment scoring.
Among many other methods that can be used to extract insights from social media sites is sentiment analysis. I believe much has been said about sentiment analysis and as you take time to read through this piece, you are probably familiar with the methods of how to extract insight using sentimental analysis methods. Just a quick mention, to classify an eWOM, you might consider the following methods (not exhaustive)
1. Topic in the text– Job advert, product advert, news, clothing
2. Type of user – Person or Organization
3. Sentiment – Positive, Neutral & Negative
4. Text content e.g.! Indicates happiness? means doubt
5. Intent in text
Having been in this industry for a while, I would like to bring in business contexts to business users about the power of social media data in forecasting sales. To start off, I feel obliged to define sentimental analysis? Sentiment analysis is a method of extracting value from unstructured textual data in order to mine people’s opinions by classifying them as either positive or negative. Opinions could emanate both from peoples public profiles or organization profiles.
In spite of the diversity of content found in online sites, predicting product sales is highly domain specific. For sales prediction to be effective, your product must receive a lot of attention which means a lot of review. Consumers are increasingly posting their opinions on social media, commenting on their experiences with products and services they purchased. To my amazement, Positive tweets by organizations do not correlated to sales.
Fortunately, positive tweets of people have more effect than by organizations. Social media activity does not affect products and services that are less socially-oriented. There is a strong evidence that social media activity has a strong effect on sales. By this, positive tweets by persons can be used to forecast sales
While doing the analysis of how online reviews can affect product or service sales, I often used to ask myself question. I believe I share my sentiments with many of you outside there.
Is there a relationship between tweets of a certain type and sales in following weeks?
Well, there seems to be a high correlation between positive tweets by persons and sales in the third and fourth week following the tweets.
To what extent the number of tweets of a certain type can be used to predict sales?
Interesting, tweets can be used to forecast sales 5 weeks after people comment positively about your product or service.
Does the high number of tweets lead to an increase in sales in the following weeks.
There is some evidence that suggests that a high number of positive personal tweets is followed by an increase in sales. During high peaks, there is a chance of reaching more positive people
Pipeline for sentiment analysis for sales people
- Identify an inbound lead (Person, Company)
- Fetch leads online comments/reviews/posts/networks
- Classify opinions based on topic, intent, type, sentiment etc.
- Pitch to customer based on their current experience or challenges
- Allow user to share opinion with other public users online to create a ripple effect
In addition to sales, there exists a growing body of knowledge on [predicting large social and economic events, in particular
- Unemployment rates
- Influenza epidemics – ability to predict peaks of influenza infection rates
- Election prediction – The fraction of attention certain parties receive on twitter corresponds to the outcome of the elections.
- Furthermore, sentiment analysis of tweets has proven useful in a financial context. One example is the study conducted by Bollen, Mao, and Zeng (2011) that uses it as a tool in order to predict the stock market. They performed a sentiment analysis on tweets about a company. The result was then compared to the share price. It was proven that a correlation between the share price and the sentiment could be found.